maintenance/rebuildlinks.inc

   1 <?
   2
   3 # Functions for rebuilding the link tracking tables; must
   4 # be included within a script that also includes the Setup.
   5 # See rebuildlinks.php, for example.
   6 #
   7
   8 # Turn this on if you've got memory to burn
   9 $wgUseMemoryTables = false;
  10
  11 # Buffer this many rows before inserting them all in one sweep. More
  12 # than about 1000 will probably not increase speed significantly on
  13 # most setups.
  14 /* private */ $rowbuf_size = 2000; // 2000 rows ~ 80 kB
  15
  16 function rebuildLinkTables()
  17 {
  18         global $wgLang, $wgUseMemoryTables, $rowbuf_size;
  19
  20         print "This script may take many hours to complete. If you abort during that time,\n";
  21         print "your wiki will be in an inconsistent state and you may have problems running\n";
  22         print "this script again. If you are going to abort, this is the time to do it.\n\n";
  23         print "Press control-c to abort (will proceed automatically in 15 seconds)\n";
  24         sleep(15);
  25
  26         $count = 0;
  27         print "Rebuilding link tables.\n";
  28
  29         print "Setting AUTOCOMMIT=1\n";
  30         wfQuery("SET SESSION AUTOCOMMIT=1", DB_WRITE);
  31
  32         print "Deleting old data in links table.\n";
  33         $sql = "DELETE FROM links";
  34         wfQuery( $sql, DB_WRITE );
  35
  36         print "Deleting old data in brokenlinks table.\n";
  37         $sql = "DELETE FROM brokenlinks";
  38         wfQuery( $sql, DB_WRITE );
  39
  40         print "Deleting old data in imagelinks table.\n";
  41         $sql = "DELETE FROM imagelinks";
  42         wfQuery( $sql, DB_WRITE );
  43
  44         print "\nAdding temporary unique index on links, brokenlinks and imagelinks.\n";
  45         print "->If build aborts now, you probably aborted a previous build. If that is\n";
  46         print "  the case, you can clean up the remains with the following SQL commands,\n";
  47         print "  and then try again.\n";
  48         print "  ALTER TABLE links DROP INDEX tmp_unique;\n";
  49         print "  ALTER TABLE brokenlinks DROP INDEX tmp_unique;\n";
  50         print "  ALTER TABLE imagelinks DROP INDEX tmp_unique;\n\n";
  51
  52         $sql = "ALTER TABLE links ADD UNIQUE tmp_unique (l_from, l_to)";
  53         wfQuery( $sql, DB_WRITE );
  54         $sql = "ALTER TABLE brokenlinks ADD UNIQUE tmp_unique (bl_from, bl_to)";
  55         wfQuery( $sql, DB_WRITE );
  56         $sql = "ALTER TABLE imagelinks ADD UNIQUE tmp_unique (il_from, il_to(244))";
  57         wfQuery( $sql, DB_WRITE );
  58         print "Temporary unique index added ok. Forget what I said.\n\n";
  59
  60         print "Locking tables\n";
  61         $sql = "LOCK TABLES cur READ, interwiki READ, user_newtalk READ, " .
  62                 "links WRITE, brokenlinks WRITE, imagelinks WRITE";
  63         wfQuery( $sql, DB_WRITE );
  64
  65         print "Finding number of articles to process\n";
  66         $sql = "SELECT COUNT(*) as count FROM cur";
  67         $res = wfQuery( $sql, DB_READ );
  68         $obj = wfFetchObject( $res );
  69         $total = $obj->count;
  70
  71         print "Finding highest article id\n";
  72         $sql = "SELECT MIN(cur_id) AS min, MAX(cur_id) AS max FROM cur";
  73         $res = wfQuery( $sql, DB_READ );
  74         $obj = wfFetchObject( $res );
  75
  76         $cur_pulser = new SelectPulser("SELECT cur_id,cur_namespace,cur_title,cur_text " .
  77                 "FROM cur WHERE cur_id ",
  78                 $obj->min, $obj->max, $rowbuf_size);
  79
  80         $brokenlinks_inserter = new InsertBuffer(
  81                 "INSERT IGNORE INTO brokenlinks (bl_from,bl_to) VALUES " , $rowbuf_size);
  82
  83         $links_inserter = new InsertBuffer(
  84                 "INSERT IGNORE INTO links (l_from,l_to) VALUES ", $rowbuf_size);
  85
  86         $imagelinks_inserter = new InsertBuffer("INSERT IGNORE INTO imagelinks ".
  87                 "(il_from,il_to) VALUES ", $rowbuf_size);
  88
  89         print "Starting processing\n";
  90
  91         $ins = $wgLang->getNsText( Namespace::getImage() );
  92         $inslen = strlen($ins)+1;
  93
  94         $tc = Title::legalChars();
  95
  96         $start_time = time();
  97         while ( $row = $cur_pulser->next() ) {
  98                 $from_id = $row->cur_id;
  99                 $ns = $wgLang->getNsText( $row->cur_namespace );
 100
 101                 $raw_title = $row->cur_title;
 102                 if ( "" != $ns ) {
 103                         $raw_title = "$ns:{$raw_title}";
 104                 }
 105                 $title = addslashes( $raw_title );
 106                 $text = $row->cur_text;
 107
 108                 $numlinks = preg_match_all( "/\\[\\[([{$tc}]+)(]|\\|)/", $text,
 109                   $m, PREG_PATTERN_ORDER );
 110
 111                 for ( $i = 0; $i < $numlinks; ++$i ) {
 112                         if( preg_match( '/^(http|https|ftp|mailto|news):/', $m[1][$i] ) ) {
 113                                 # an URL link; not for us!
 114                                 continue;
 115                         }
 116
 117                         # FIXME: Handle subpage links
 118                         $nt = Title::newFromText( $m[1][$i] );
 119
 120                         if (! $nt)
 121                         {
 122                                 $txt = $m[1][$i];
 123                                 print "error in '$ns:{$row->cur_title}' :\t'$txt'\n";
 124                                 continue;
 125                         }
 126                         if( $nt->getInterwiki() != "" ) {
 127                                 # Interwiki links are not stored in the link tables
 128                                 continue;
 129                         }
 130                         if( $nt->getNamespace() == Namespace::getSpecial() ) {
 131                                 # Special links not stored in link tables
 132                                 continue;
 133                         }
 134                         if( $nt->getNamespace() == Namespace::getMedia() ) {
 135                                 # treat media: links as image: links
 136                                 $nt = Title::makeTitle( Namespace::getImage(), $nt->getDBkey() );
 137                         }
 138
 139                         $dest = addslashes( $nt->getPrefixedDBkey() );
 140                         $dest_id = $nt->getArticleID();
 141
 142                         if ( 0 == strncmp( "$ins:", $raw_title, $inslen ) ) {
 143                                 $iname = addslashes( substr( $raw_title, $inslen ) );
 144                                 $imagelinks_inserter->insert( "('{$title}','{$iname}')" );
 145                         } else if ( 0 == $dest_id ) {
 146                                 $brokenlinks_inserter->insert( "({$from_id},'{$dest}')" );
 147                         } else {
 148                                 $links_inserter->insert( "('{$title}',{$dest_id})" );
 149                         }
 150                 }
 151
 152                 if ( ( $count % 10 ) == 0 )
 153                         print ".";
 154
 155                 if ( ( ++$count % 1000 ) == 0 ) {
 156                         $dt = time() - $start_time;
 157                         $start_time = time();
 158                         $rps = ($dt == 0 ? "lots of" : intval(1000/$dt));
 159                         print "\n$count of $total articles scanned ({$rps} articles per second)\n";
 160                 }
 161
 162         }
 163
 164         $imagelinks_inserter->flush();
 165         $links_inserter->flush();
 166         $brokenlinks_inserter->flush();
 167
 168         print "$total articles scanned.\n";
 169
 170         print "Removing temporary unique indexes from tables links, brokenlinks and imagelinks.\n";
 171         $sql = "ALTER TABLE links DROP INDEX tmp_unique";
 172         wfQuery( $sql, DB_WRITE );
 173         $sql = "ALTER TABLE brokenlinks DROP INDEX tmp_unique";
 174         wfQuery( $sql, DB_WRITE );
 175         $sql = "ALTER TABLE imagelinks DROP INDEX tmp_unique";
 176         wfQuery( $sql, DB_WRITE );
 177
 178         $sql = "UNLOCK TABLES";
 179         wfQuery( $sql, DB_WRITE );
 180         print "Done\n";
 181 }
 182
 183 # InsertBuffer increases performance slightly by inserting many rows
 184 # at once. The gain is small (<5%) when running against a local, idle
 185 # database, but may be significant in other circumstances. It also
 186 # limits the number of inserted rows uppwards, which should avoid
 187 # problems with huge articles and certain mysql settings that limits
 188 # the size of queries. It's also convenient.
 189
 190 class InsertBuffer {
 191         /* private */ var $mBuf, $mSql, $mBufcount, $mMaxsize;
 192
 193         function InsertBuffer( $sql, $bufsize ){
 194                 $this->mSql = $sql;
 195                 $this->mBuf = array();
 196                 $this->mBufcount = 0;
 197                 $this->mMaxsize = $bufsize;
 198         }
 199
 200         function insert( $value ){
 201                 // print $this->mSql . " -> " . $value . "\n";
 202                 $this->mBuf[] = $value;
 203                 $this->mBufcount++;
 204                 if($this->mBufcount > $this->mMaxsize){
 205                         $this->flush();
 206                 }
 207         }
 208
 209         function flush(){
 210                 if( $this->mBufcount > 0 ){
 211                         $sql = $this->mSql . implode(",", $this->mBuf);
 212                         wfQuery( $sql, DB_WRITE );
 213                         $this->mBuf = array();
 214                         $this->mBufcount = 0;
 215                         // print "Wrote query of size " . strlen( $sql ) . "\n";
 216                 }
 217         }
 218
 219 }
 220
 221 # Select parts from a large table by using the "BETWEEN X AND Y"
 222 # operator on the id column. Avoids buffering the whole thing in
 223 # RAM. It's also convenient.
 224 class SelectPulser {
 225         /* private */ var $mSql, $mSetsize, $mPos, $mMax, $mSet;
 226
 227         function SelectPulser( $sql, $min, $max, $setsize) {
 228                 $this->mSql = $sql;
 229                 $this->mSet = array();
 230                 $this->mPos = $min;
 231                 $this->mMax = $max;
 232                 $this->mSetsize = $setsize;
 233         }
 234
 235         function next(){
 236                 $result = current( $this->mSet );
 237                 next( $this->mSet );
 238                 if( false !== $result ){
 239                         return $result;
 240                 }
 241                 while( $this->mPos <= $this->mMax ){
 242                         $this->mSet = array();
 243                         $sql = $this->mSql . " BETWEEN " . $this->mPos .
 244                                 " AND " . ($this->mPos + $this->mSetsize - 1);
 245                         $this->mPos += $this->mSetsize;
 246
 247                         $res = wfQuery( $sql, DB_READ );
 248                         while ( $row = wfFetchObject( $res ) ) {
 249                                 $this->mSet[] = $row;
 250                         }
 251
 252                         if( count( $this->mSet ) > 0 ){
 253                                 return $this->next();
 254                         }
 255                 }
 256                 return false;
 257         }
 258 }
 259
 260 ?>